Mixed-flow water turbine vibration signal processing method based on adaptive enhanced EEMD (ensemble empirical mode decomposition)

Abstract

The invention discloses a mixed-flow water turbine vibration signal processing method based on adaptive enhanced EEMD (ensemble empirical mode decomposition), and relates to the technical field of vibration signal analysis and processing. A vibration signal of the mixed-flow water turbine is collected, and wavelet noise reduction is carried out; self-adaptive frequency band weighted noise is injected into the noise-reduced signal, and a noise-added signal is obtained; eMD decomposition is carried out on the signal after noise adding, a stopping condition is modified into a double-constraint mode stopping criterion in the EMD decomposition process, and an IMF component set and a residual error are obtained by using double constraints of the number of extreme points and the residual error energy change rate; and effectively screening IMF components of the IMF based on a fault sensitivity index, and introducing a compensation item into a residual error to obtain a final decomposition result as a vibration signal processing result. Background noise is effectively suppressed through frequency band weighted noise injection, and modal aliasing is effectively suppressed through a double-constraint stop criterion and fault sensitivity screening, so that fault features can be separated more accurately, and the signal-to-noise ratio can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vibration signal analysis and processing, and in particular to a vibration signal processing method for a Francis turbine based on an adaptive enhanced EEMD. Background Art

[0002] Water in a Francis turbine enters the runner radially and exits axially. This design results in a compact structure and stable operation. It also boasts high conversion efficiency, maintaining high power generation efficiency under varying head and flow conditions. This makes Francis turbines widely applicable to hydropower plants of all sizes. Francis turbine vibration signals often contain a significant amount of abnormal data, so preprocessing is typically required when assessing and predicting unit health based on these vibration signals.

[0003] Ensemble empirical mode decomposition (EEMD) builds on EMD decomposition by repeatedly adding white noise to improve decomposition stability and mitigate modal aliasing. It leverages the flat spectrum of white noise and offsets the effects of noise on IMF components through repetition and averaging, thereby highlighting the true characteristics of the signal. Due to its excellent properties, it is often used to process nonlinear and non-stationary signals.

[0004] However, when processing Francis turbine vibration signals, especially those in fault conditions, there are still problems of insufficient signal decomposition and insufficient adaptability. Summary of the Invention

[0005] The purpose of the present invention is to provide a Francis turbine vibration signal processing method based on adaptive enhanced EEMD to solve the problems of insufficient signal decomposition and insufficient adaptability in existing EEMD decomposition processing.

[0006] To solve the above technical problems, the present invention adopts the following technical solution: a method for processing vibration signals of a Francis turbine based on an adaptive enhanced EEMD, characterized by comprising the following steps: S1. Collect vibration signals of Francis turbine , and perform wavelet denoising to obtain ; S2. injecting adaptive frequency-band weighted noise into the denoised signal to obtain a noisy signal; S3. Perform EMD decomposition on the noisy signal. During the EMD decomposition process, modify the stopping condition to a double-constrained modal stopping criterion, using the dual constraints of the number of extreme points and the rate of change of residual energy to obtain the IMF component set and residual; S4. Effectively screen the IMF components based on the fault sensitivity index and introduce compensation terms into the residual to obtain the final decomposition result as the vibration signal processing result.

[0007] A further technical solution is that the specific steps of step S2 are as follows: S2-1. Generate noise according to the following formula : in, It is the sub-band after the signal frequency band is divided; Frequency band The noise gain factor, ; is the frequency band weight function, which controls the distribution of noise in different frequency bands and is dynamically calculated by the power spectrum density; For the Gaussian white noise of frequency bands, with a mean of 0 and a standard deviation of ; The calculation of is as follows: in, For signals in the frequency band The power spectral density of is obtained by short-time Fourier transform; is the power threshold, used to distinguish high / low power frequency bands, and takes 10% to 20% of the total signal power, that is, ; is the steepness parameter of the Sigmoid function, ;in in, , Indicates vibration signal The standard deviation of Indicates the total signal power; S2-2. Generate the noisy signal: .

[0008] Indicates the number of decompositions.

[0009] A further technical solution is that the dual-constraint modal stopping criterion includes: in, is the control coefficient of the number of extreme points, ; is the signal length; is the residual energy change rate threshold, which is used to determine whether the decomposition has converged. , applicable to non-stationary signals; For the The second decomposition The residual signal after iterations.

[0010] A further technical solution is that the applicable conditions of the dual-constraint modal stopping criterion are as follows: When the signal frequency band is greater than 500Hz, , ; When the signal frequency is less than 10Hz, , ; When the signal frequency range is 10Hz~500Hz, , .

[0011] A further technical solution is that the specific steps of step S4 are as follows: S4-1. Define fault sensitivity index : in, is the energy ratio, is the frequency band enhancement factor, For the The energy of the IMF, ; is the total energy of all IMFs; is a typical frequency set of common faults of hydropower units, is the IMF component at frequency The amplitude at is the average spectrum amplitude of the IMF component; S4-2. Filter effective IMF using the following formula satisfy The retention of , those that do not meet the requirement are counted as residuals, represents the fault sensitivity threshold; S4-3. A compensation term is introduced into the residual to enhance the sparsity of high-frequency impulse components through the sign function and suppress low-frequency modulation interference. The final residual formula is as follows: in, For the The compensation weight coefficient of each IMF is obtained by normalizing the kurtosis value; is a sign function, and the IMF derivative is taken; S4-4. Obtain the final decomposition result as the vibration signal processing result, the formula is as follows: .

[0012] Compared with the existing technology, the beneficial effects of the present invention are: effective suppression of background noise through frequency band weighted noise injection, effective suppression of modal aliasing through dual-constraint stopping criteria and fault sensitivity screening, so that fault characteristics can be more accurately separated and the signal-to-noise ratio can be improved. Its adaptive noise strategy and fault sensitivity screening mechanism can dynamically balance noise suppression and feature retention, so that the decomposed signal has both high resolution and strong robustness in the time-frequency domain. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 Schematic diagram of high-speed acquisition system definition in the embodiment.

[0014] Figure 2 2 is a sampling waveform diagram in the embodiment.

[0015] Figure 3 It is a waveform diagram after decomposition in the embodiment.

[0016] Figure 4 2 is a time-frequency analysis comparison diagram in the embodiment.

[0017] Figure 5 2 is a comparison diagram of envelope spectra in the embodiments.

[0018] Figure 6 2 is a comparison chart of impact energy retention rates in the examples. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] Example The decomposition process of traditional EEMD can be summarized as follows: Noise signal generation: in, It is White noise is added, is a fixed noise standard deviation, and M is the number of decompositions.

[0021] Modal decomposition For each Perform EMD decomposition to obtain the IMF set and residuals Ensemble mean The final IMF is the mean of all decomposition results: To address the problem that the above methods are insufficient for decomposing Francis turbine vibration signals, the present invention proposes a fault-sensitive adaptive enhanced EEMD (AE-FS-EEMD) decomposition method for processing Francis turbine vibration signals, comprising the following steps: S1. Collect vibration signals of Francis turbine , and perform wavelet denoising to obtain .

[0022] S2. Inject adaptive frequency-weighted noise into the denoised signal to obtain a noisy signal. Reduce noise injection in high-power frequency bands (such as the fault characteristic frequency) to avoid masking the characteristic; enhance noise in low-SNR frequency bands to improve modal separation stability.

[0023] The specific steps are as follows: S2-1. Generate noise according to the following formula : in, It is the sub-frequency band after the signal frequency band is divided. The specific value is the multiple of the rotation frequency. For example, the rotation frequency harmonics such as shaft misalignment and rotor imbalance are generally 2 times the frequency, so k is 2; Frequency band The noise gain factor, ; is the frequency band weight function, which controls the distribution of noise in different frequency bands and is dynamically calculated by the power spectrum density; For the Gaussian white noise of frequency bands, with a mean of 0 and a standard deviation of ; The calculation of is as follows: in, For signals in the frequency band The power spectral density of is obtained by short-time Fourier transform; is the power threshold used to distinguish high / low power frequency bands, and takes 10% to 20% of the total signal power (when the signal is a strong periodic signal, take 10%; when in a strong noise environment, such as cavitation conditions, take 20%; in other cases, take 15%), that is, ; is the steepness parameter of the Sigmoid function, ;in in, , Indicates vibration signal The standard deviation of Indicates the total signal power.

[0024] S2-2. Generate the noisy signal: .

[0025] S3. Perform EMD decomposition on the noisy signal. During the EMD decomposition process, modify the stopping condition to a dual-constrained modal stopping criterion, using the dual constraints of the number of extreme points and the rate of change of the residual energy to obtain the IMF component set and residual. The dual-constrained modal stopping criterion avoids premature termination when decomposing the high-frequency impact components of the hydropower unit, ensuring that weak fault characteristics are not missed.

[0026] The dual-constraint modal stopping criteria include: in, is the control coefficient of the number of extreme points, ; is the signal length; is the residual energy change rate threshold, which is used to determine whether the decomposition has converged. , applicable to non-stationary signals; For the The second decomposition The residual signal after iterations.

[0027] The above dual-constraint modal stopping criteria are applicable under the following conditions: When the signal frequency band is greater than 500Hz, , When decomposing high-frequency components (such as bearing impact and gear meshing noise), the signal changes dramatically instantaneously and the number of extreme points may decrease rapidly, but weak impact features are easily masked by noise.

[0028] When the signal frequency is less than 10Hz, , When decomposing low-frequency components (such as hydraulic imbalance and wheel swing), the signal changes smoothly and the residual energy change rate may be lower than the threshold for a long time, but the number of extreme points reflects the physical meaning of the mode.

[0029] When the signal frequency range is 10Hz~500Hz, , When decomposing intermediate frequency components (such as mechanical looseness and misalignment), the signal exhibits both non-stationary and periodic characteristics, as shown in Table 1.

[0030] Table 1 Parameter setting reference table The specific Python code (part) is as follows: def stop_criterion(r_prev, r_current, extrema_count, imf_index): N = len(r_prev) gamma = 1.2 if imf_index <= 3 else (0.6 if imf_index >= 4 else0.8) epsilon = 0.01 if imf_index <= 3 else (0.05 if imf_index >= 4else 0.02) # Calculate the residual energy change rate energy_ratio = np.linalg.norm(r_current - r_prev) / np.linalg.norm(r_prev) # Judgment conditions if imf_index <= 3: # high frequency stage return energy_ratio <= epsilon # only depends on criterion 2 elif imf_index >= 4: # low frequency stage return extrema_count <= gamma * np.log(N) # Depends only on criterion 1 else: #Intermediate frequency stage return (extrema_count <= gamma * np.log(N)) and (energy_ratio<= epsilon) S4. Effectively screen the IMF components based on the fault sensitivity index and introduce compensation terms into the residual to obtain the final decomposition result as the vibration signal processing result.

[0031] The specific steps of step S4 are as follows: S4-1. Define fault sensitivity index : in, is the energy ratio, is the frequency band enhancement factor, For the The energy of the IMF, ; is the total energy of all IMFs; is a typical frequency set of common faults of hydropower units, is the IMF component at frequency The amplitude at is the average spectrum amplitude of the IMF component; S4-2. Filter effective IMF using the following formula satisfy The retention of , those that do not meet the requirement are counted as residuals, represents the fault sensitivity threshold; S4-3. A compensation term is introduced into the residual to enhance the sparsity of high-frequency impulse components through the sign function and suppress low-frequency modulation interference. The final residual formula is as follows: in, For the The compensation weight coefficient of each IMF is obtained by normalizing the kurtosis value; is a sign function, and the IMF derivative is taken; S4-4. Obtain the final decomposition result as the vibration signal processing result, the formula is as follows: .

[0032] To verify the above method, The data acquisition system hardware mainly consists of three parts: sensors, acquisition cards, and edge computers. The vibration sensor is MLS-9, the sound sensor is AWA14423, the eddy current sensor is RSW3308, and the pressure pulsation sensor is AK-4C. The acquisition cards are Advantech's IDAQ-801 series and IDAQ-817 series. Similarly, the edge computer is Advantech's UNO-348 series. The specific equipment parameters are shown in Table 2: Table 2 Hardware configuration The high-speed acquisition module uses Advantech's USB high-speed acquisition series board with the corresponding data acquisition driver. The vibration signals collected by this system are all AC-coupled voltage signals. The basic information of the high-speed acquisition system sensor is defined as follows: Figure 1 As shown in (a), this block diagram mainly determines the channel position, amplitude range, coupling mode, etc. of the sensor used; the vibration signals collected by this system are all AC coupled-voltage signals. The acquisition task definition is as follows Figure 1As shown in (b), in the middle of the figure, the number of sampling points, sampling rate and sampling mode need to be configured according to the type of sensor connected to different acquisition cards. In this system, the sampling mode selects multi-channel multi-sampling (multi-channel means that all channels used by the acquisition card can work simultaneously, and multi-sampling means collecting multiple points at a time. Generally, the number of multi-sampling points is the same as the sampling rate, which means the amount of data collected once per second. Multi-sampling facilitates fast Fourier transform). The sampling rate and sampling point number of the high-speed acquisition card equipped with the sound sensor are both 20k, and the sampling rate and sampling point number of other acquisition cards are both 2048.

[0033] In order to fit the theme of this article, but limited by the limitations of real fault data. A destructive experiment was carried out using an abandoned runner. A notch was artificially cut in the runner blades and then installed on the original unit to simulate the wear failure of the runner during operation. Taking safety factors into consideration, this experiment only retains the data extracted for 10 seconds after stable operation. The experiment was carried out on a unit in the Hydro-Mechanical Coupling Laboratory of Kunming University of Science and Technology. The unit is a mixed flow turbine unit manufactured by Yunnan Yuxi Hydropower Equipment Company, model HLA551-LJ-43, with a runner diameter of 43 cm, a designed head of 12 m, a rated flow of 0.7 m3 / s, and a turbine output of 65 kW; the generator is SF55-10 / 740, with a rated capacity of 55 kW and a speed of 600 rpm. The unit maintains a complete unit shaft system and support structure, which can be used to study the unit shaft system vibration characteristics. The test unit is designed according to the actual production requirements of the power plant and can achieve grid-connected operation and operation with isolated load. The specific experimental waveform is as follows Figure 2 shown.

[0034] The adaptive enhanced EEMD operation is performed on the vibration fault sequence to obtain the IMFs as follows: Figure 3 As shown in the figure, in terms of the selection of the number of sets and the decomposition noise intensity, as the number of sets increases, the influence of the Gaussian white noise added in the adaptive enhanced EEMD process on the decomposition effect gradually decreases and tends to be stable. Different studies have shown that the influence of noise intensity on the result error is also relatively slight. Therefore, when using the adaptive enhanced EEMD to process the vibration fault time series, the number of sets is set to 200 and the auxiliary noise intensity is set to 0.4.

[0035] To further verify the effectiveness and practicality of the method proposed in the present invention, a comparison of various methods was conducted. The details are as follows: Time-frequency analysis comparison The comparison results are shown in Figure 4. From the time-frequency analysis results, it can be seen that the adaptive enhanced EEMD method shows significant advantages in decomposing the unit vibration signal: its IMF3 component shows a high-energy concentration area at the target fault frequency, with a compact energy distribution and clear boundaries, indicating that this method can accurately extract high-frequency impact characteristics, while the IMF3 of the traditional EEMD has dispersed energy and mixed low-frequency components in the same frequency band, resulting in modal aliasing; at the same time, the adaptive enhanced EEMD effectively suppresses high-frequency noise through wavelet pre-denoising and frequency band weighted noise injection, and the energy base is significantly lower than the traditional method (non-characteristic frequency bands are uniformly blue), while the traditional EEMD has sporadic high-energy points in the same area due to the fixed noise strategy, indicating that the noise residue is serious. This improvement is due to the dual-constraint stopping criterion's retention of high-frequency impact components and dynamic suppression of low-frequency interference, which makes the adaptive enhanced EEMD have both high resolution and strong robustness in the time-frequency domain.

[0036] Envelope spectrum comparison chart The comparison results are shown in Figure 5 . As can be seen from the envelope spectrum comparison diagram, the adaptive enhanced EEMD method has significant advantages in feature extraction and noise suppression: in the target frequency band of 100-200Hz, the amplitude of the adaptive enhanced EEMD is significantly higher than that of VMD and traditional EEMD. For example, at 150Hz, the amplitude of the adaptive enhanced EEMD is about -20dB, while the traditional EEMD is lower than -40dB, indicating that it retains the energy of the relevant frequencies more completely; at the same time, the amplitude of the adaptive enhanced EEMD in the non-fault frequency band below 100Hz is generally lower, with a peak value close to -80dB, verifying its effective suppression of background noise through frequency band weighted noise injection; and reflecting the dual-constraint stopping criterion and fault sensitivity screening to suppress modal aliasing. This result proves that the adaptive enhanced EEMD can more accurately separate fault features and improve the signal-to-noise ratio.

[0037] Comparison of impact energy retention rate (different SNR) The comparison results are shown in Figure 6The impact energy retention rate comparison chart clearly shows that the adaptive enhanced EEMD method significantly outperforms traditional EEMD in robustness and signal retention under low signal-to-noise ratio (SNR) conditions. Under strong noise interference at an SNR of -5dB, the adaptive enhanced EEMD still retains approximately 90% of the impact energy, while the traditional EEMD retains less than 50%. This demonstrates that wavelet pre-noise reduction and band-weighted noise injection effectively suppress noise from drowning out key features. As the SNR increases to 10dB, the adaptive enhanced EEMD's retention rate reaches 95%, while the traditional EEMD only reaches 70%. This verifies the ability of its dual-constrained stopping criterion to accurately capture high-frequency, weak impact components. Furthermore, the overall trend of the curve shows that the adaptive enhanced EEMD's retention rate consistently outperforms traditional methods at varying SNRs. This advantage is particularly pronounced at extremely low SNRs (SNR < 0dB), where the retention rate increases by over 45%. This demonstrates that its adaptive noise strategy and fault sensitivity screening mechanism can dynamically balance noise suppression and feature retention.

[0038] Although the invention has been described herein with reference to a number of illustrative embodiments thereof, it should be understood that numerous other modifications and embodiments can be devised by those skilled in the art that will fall within the scope of this disclosure.

Claims

1. A vibration signal processing method for a Francis turbine based on adaptive enhanced EEMD, characterized in that The steps include: S1. Collect vibration signals of Francis turbine , and perform wavelet denoising to obtain ; S2. injecting adaptive frequency-band weighted noise into the denoised signal to obtain a noisy signal; S3. Perform EMD decomposition on the noisy signal. During the EMD decomposition process, modify the stopping condition to a double-constrained modal stopping criterion, using the dual constraints of the number of extreme points and the rate of change of residual energy to obtain the IMF component set and residual; S4. Effectively screen the IMF components based on the fault sensitivity index and introduce compensation terms into the residual to obtain the final decomposition result as the vibration signal processing result.

2. The method according to claim 1, wherein: The specific steps of step S2 are as follows: S2-1. Generate noise according to the following formula : in, It is the sub-band after the signal frequency band is divided; Frequency band The noise gain factor, ; is the frequency band weight function, which controls the distribution of noise in different frequency bands and is dynamically calculated by the power spectrum density; For the Gaussian white noise of frequency bands, with a mean of 0 and a standard deviation of ; The calculation of is as follows: in, For signals in the frequency band The power spectral density of is obtained by short-time Fourier transform; is the power threshold, used to distinguish high / low power frequency bands, and takes 10% to 20% of the total signal power, that is, ; is the steepness parameter of the Sigmoid function, ;in in, , Indicates vibration signal The standard deviation of Indicates the total power of the vibration signal; S2-2. Generate the noisy signal: Indicates the number of decompositions.

3. The method according to claim 1, wherein: The dual-constraint modal stopping criteria include: in, is the control coefficient of the number of extreme points, ; is the signal length; is the residual energy change rate threshold, which is used to determine whether the decomposition has converged. , applicable to non-stationary signals; For the The second decomposition The residual signal after iterations.

4. The method according to claim 3, wherein: The applicable conditions of the dual-constraint modal stopping criterion are as follows: When the signal frequency band is greater than 500Hz, , ; When the signal frequency is less than 10Hz, , ; When the signal frequency range is 10Hz~500Hz, , .

5. The method according to claim 1, wherein: The specific steps of step S4 are as follows: S4-1. Define fault sensitivity index : in, is the energy ratio, is the frequency band enhancement factor, For the The energy of the IMF, ; is the total energy of all IMFs; is a typical frequency set of common faults of hydropower units, is the IMF component at frequency The amplitude at is the average spectrum amplitude of the IMF component; S4-2. Filter effective IMF using the following formula satisfy The retention of , those that do not meet the requirement are counted as residuals, represents the fault sensitivity threshold; S4-3. A compensation term is introduced into the residual to enhance the sparsity of high-frequency impulse components through the sign function and suppress low-frequency modulation interference. The final residual formula is as follows: in, For the The compensation weight coefficient of each IMF is obtained by normalizing the kurtosis value; is a sign function, and the IMF derivative is taken; S4-4. Obtain the final decomposition result as the vibration signal processing result, the formula is as follows: 。

Images